Support vector machines for candidate nodules classification

Image processing techniques have proved to be effective for the improvement of radiologists' diagnosis of lung nodules. In this paper, we present a computerized system aimed at lung nodules detection; it employs two different multi-scale schemes to identify the lung field and then extract a set of candidate regions with a high sensitivity ratio. The main focus of this work is the classification of the elements in the very unbalanced candidates set, by the use of support vector machines (SVMs). We performed several experiments with different kernels and differently balanced training sets. The results obtained show that cost-sensitive SVMs trained with very unbalanced data sets achieve promising results in terms of sensitivity and specificity.